Testing the Creativity of Randomness

Can random sequences of an alphabet convey meaning? Can random sequences of amino acids perform a function? A new study tested a subset of random sets of amino acids expanded with tandem duplications. The results were sometimes surprising but posed no threat to intelligent design.

Douglas Axe’s groundbreaking work into the dark matter of protein conformational space in 2004 strongly suggested that functional space represents a vanishingly small fraction of sequence space.¹ He mutated a beta lactamase enzyme to the point where it would no longer fold into its functional structure and estimated that the ratio of any functional protein that folds correctly to random sequences of 150 amino acids was roughly 1 in 10⁷⁴. This implies that the vast field of possible sequences represents useless dark matter, comparable to strings of random letters happening, on extremely rare occasions, to spell meaningful words or sentences. The longer the meaningful sequence, the more improbable it becomes, exponentially.²

Evolutionary scientists recognize this needle-in-a-haystack problem. In PNAS, a team of researchers from UC San Francisco along with colleagues at Brandeis University state,

Proteins are molecular machines that rely on well-defined tertiary structures to carry out many of their remarkable functions. The molecular mechanisms by which proteins evolved their modern-day sequences and structures have been a puzzle for decades. Early studies focused on the astronomical number of possible sequences for even a 100-residue protein there are 20¹⁰⁰ possible permutations for a sequence of this length composed of the 20 standard amino acids.

That’s equivalent to 1.3 x 10¹³⁰ trials evolution would have to search through to locate functional protein folds. To get around this formidable challenge to unguided chemical evolution, the authors speculate that proteins may have evolved from shorter peptides with simple folds. They also point out that multiple sequences could arrive at the same fold; they define this as “designability.” Their paper is titled, “The emergence of novel versus known three-dimensional structures from random sequences.”

To make the search for functional polypeptides within the vast arena of sequence space more tractable, they consider the possibility that tandem repeats of polypeptides would be more amenable to folding. Among the folding sequences — a small corner of the arena — functional polypeptides would then be more easily found. We could analogize tandem repeats as short metal bars with hinges that might fold into random 3D structures, a few of which might be useful.

They tested this hypothesis primarily in silico, with some follow-up corroboration using selected promising sequences inserted into the E. coli genome. This was to see if the predicted folds actually did fold in vivo. They generated populations of a million samples of sequences 5 to 60 amino acids long, repeated enough times to generate polypeptides 120 units long. To make it even simpler for chemical evolution to get a head start, they also checked random sequences using reduced alphabets of 10 amino acids — half the number of canonical amino acids found in living cells. “Similar trends were observed for both alphabets up to 40-residue repeats,” they claim.

Together, these data show that there are relatively few designable conformations for structures with short repeat lengths (Fig. 1H), but this number increases as the number of mainchain torsional degrees of freedom increases. However, at repeat lengths longer than 30 residues, foldable sequences occur with such low frequencies that we observe far fewer within our sample size of one million sequences/repeat length.

Speaking as optimistically as their evolutionary hopes permitted, they did find some squares, triangles, ovals, helices, and other structures that software predicted were foldable from random strings of 6, 8, and 10 amino acids extended with tandem repeats. The most common structures were solenoids (pipe-like helices) and amyloids (beta sheets), some of which resembled domains in natural proteins. Testing longer sequences soon outpaced their computational resources. To sample that space, they selected folding structures within their million samples of each n-mer (a tiny subset of potential polypeptides).

They became particularly excited about an octamer that formed a novel helix they named HeliScrew4 and a novel supersecondary structure they named the “alpha-helical screw” that emerged.

Hopeful Monsters

Their work can be appreciated for the part about investigating novel structures for possible applications to protein design. (Notice the sales pitch tacked onto the materialist dogma: “These findings advance our understanding of molecular evolution and have implications for design of novel proteins.” The latter purpose does not justify the former.)

It should not be that surprising that repetition generates patterns. A string of peptides that coils by natural forces might grow indefinitely when repeated, like a coil of razor wire. If bounded by introduced breaks, it might turn corners. Let’s think, though, about their basic premise: repeating strings can be a source of functional information. Is there creative power in random sequences repeated multiple times to create longer strings? Consider this 4-mer repeated 8 times:

dumbdumbdumbdumbdumbdumbdumbdumb

The sequence did not get smarter by being repeated. The authors suggest that indels (insertions and deletions) might free up the powers of evolution. After all, natural proteins often do contain domains of repeated sequences.

These calculations suggested that intramolecularly folded helical bundles might be formed with high frequency only if one allowed occasional breaks between blocks of heptad repeats to allow the protein helices to fold back on one another. To test this idea, we systematically introduced INDELs, consisting of either 1- or 2-residue deletions or 1 to 4-residue Gly insertions between blocks of four repeated random heptads (to create app. 28-residue helices). This pattern was repeated four times to create potential 4-helix bundles. The presence of deletions indeed increased the fraction of sequences that folded with high confidence by 20-fold, from 0.002% to approximately 0.04%. More dramatic, 300 to 500-fold increases were seen with 1 to 4-residue Gly insertions, reaching 0.93% for a two-residue insertion (Fig. 8A). The large majority (39.7%) of the predicted structures represented classical left or right-turning helical bundles….

Notice the investigator interference, “we systematically introduced” indels. Were there mutations selected by intelligent design? Would unguided natural processes be likely to make these discoveries without intelligent help?³

Since indels are said to exhibit creativity, let’s try an insertion and a deletion in our test string:

dumbdumbdumdumbdumbdumbdumbdumber

A hexamer like “stupid” would be unlikely to fare better, nor a repeated 10-mer like “nincompoop.”

Ah, they think, but there would be “selective pressure” for getting smarter:

Gene duplication resulted in repeated sequences in which each individual repeat could evolve independently and asymmetrically under selective pressure for function.

As I stated in an earlier article, selective pressure is a false force. The easiest job for evolution is extinction. Thinking that “selective pressure” causes higher function to emerge begs the question of whether natural selection possesses intrinsic creative power.⁵

The evolutionary part of this paper relies on wishful thinking. What “could” or “might” happen occurs only in their evolutionary imaginations. In one of the team’s randomly constructed solenoids, they identified potential binding sites for Ca²⁺ ions that they imagined might suggest a way to explain the origin of metalloproteins.

…our workflow identified many very reasonable candidates for metal ion-binding motifs….

We also observe clusters of His, Glu/Asp, and Cys residues that could bind transition metals known to catalyze hydrolytic and oxidative reactions.

How a cell would handle a toxic metallic ion floating in its midst without the assistance of chaperones was not addressed.

Reality Check

After I read the entire paper, an AI engine I chatted with (Gemini) agreed with significant problems I identified. I pointed out instances of investigator interference, unrealistic conditions, and unwarranted extrapolations. It confirmed my suspicions that the paper, overall, was a big, unrealistic bluff.

Initially, Gemini (which knew all about Douglas Axe’s research) responded that a 15-mer repeated 10 times would lower the probability substantially from Axe’s estimate. Perhaps folds in a 150-unit sequence containing repeated random strings could emerge in 1 in 10¹⁹ trials. 10¹⁹ is a large number, for sure, but is 58 orders of magnitudes smaller. It might make a random search more tractable given the probabilistic resources of the early earth. Echoing the consensus of scientific materialists, Gemini drew on other evolutionists’ calculations:

Studies regularly find that simple, weak binding or basic catalytic activity can occur in random libraries at rates closer to 1 in 10¹¹ or 1 in 10¹⁴.

While 1 in 10¹⁴ is still a rare event, it is 60 orders of magnitude more common than 10⁷⁷. In a single test tube filled with quadrillions of random peptides, a 1-in-10¹⁴ event is practically guaranteed to happen multiple times. Mainstream biology relies on these empirical screen results rather than Axe’s theoretical mathematical model.

But there’s a catch that I noticed and Gemini confirmed: the authors of the current study, and these others, began with 100 percent homochiral amino acids. The improbability skyrockets when evolution has to deal with racemic mixtures that would likely prevail on the early earth. Axe’s calculation goes from 1 in 10⁷⁴ to 1 in 10¹⁴⁸ (actually, 1 in 10¹⁶⁴ when requiring only peptide bonds).⁴ The optimistic 1 in 10¹⁴  estimate deflates to less than 1 in 10²⁸, a billion trillion trillion trillion trillion times less probable on average. If a 1 in 10²⁸ probability still sounds tractable to a chemical evolutionist, consider that if the lucky structure must interact functionally with other lucky structures, the likelihood quickly outstrips the probabilistic resources of the earth or even the universe. The evidence for design remains Undeniable.

Here are other problems with the paper that make its evolutionary speculations unrealistic.

• A sequence that folds in software might not fold in real life. Cells are messy places with numerous competing cross reactions. Without chaperones, a “possible” fold might not translate to an actual fold. In fact, these researchers found that out experimentally when trying out their sequences in E. coli (see quote in footnote 3).
• If a folded sequence occurred in a protocell or in the ocean, it would be subject to water attack (hydrolysis) and would be unlikely to persist for long.
• Solenoids and amyloids function only as domains within larger proteins and systems. In a cell they are likely to quickly degrade into sludge. Amyloids are notorious for this propensity, being implicated in diseases like Alzheimer’s. And a solenoid without an accompanying piston won’t ring the doorbell.
• A potentially functional structure is useless on its own. Proteins in life participate in an “interactome” of partners that work together within biological condensates. Gemini dubbed this a “Systems Biology Chasm” for chemical evolution to cross. I like how it characterized this issue:

Without this pre-existing coordination, a single, highly stable 120-mer is just molecular clutter. In fact, if it has sticky, uncoordinated surfaces, it is worse than useless — it will bind to things randomly, disrupting other fragile prebiotic reactions.

• Meaning emerges not from repetition, but from specified complexity. As in language, it is the aperiodicity of a sequence of letters — its asymmetric sequence tied to immaterial language conventions — where understanding is located. Repetition can be used for effect (“A rose is a rose is a rose”) but the exalted thoughts in a document like the Declaration of Independence reside in its semantic aperiodicity carefully arranged by a mind.
• In a protein, the active site is unlikely to reside in a tandem repeat.
• Looking for stable folds ignores the trending concept of Intrinsically Disordered Proteins (read about this here) that do not fold into stable structures yet appear to entail multiple functions due to their flexibility.
• Even if, through the most unlikely scenario, a random string of amino acids folded into a potentially useful structure, how would that information become back-coded into DNA? It would be the end of the line, like a wrench on the ground with no one knowing where it came from and how and where to use it. What of the promoters, start/stop codons and ribosome binding sites that would also be required to translate the code into a 3-D printed wrench? Gemini called this reverse-information problem “The Ultimate Catch-22.”

As is so often the case with research into chemical evolution, this paper, despite its sophisticated methods, ends up guilty of cheating (starting with pure homochiral chemicals), investigator interference, hope and dogma. Its attempts to justify the effort by promising “understanding” and teasing “implications for design of novel proteins”⁴ sounds like what Karl Popper dubbed “promissory materialism” — the tendency to hide materialistic explanations in the future.⁵ For protein design that might advance healthcare or nanotechnology, the research is welcome. For chemical evolution, it amounts to one big bluff.

Notes

1. Douglas D. Axe, “Estimating the prevalence of protein sequences adopting functional enzyme folds,” Journal of Molecular Biology, 2004 Aug 27;341(5):1295-315. DOI:10.1016/j.jmb.2004.06.058. “…the overall prevalence of sequences performing a specific function by any domain-sized fold may be as low as 1 in 10(77), adding to the body of evidence that functional folds require highly extraordinary sequences.”
2. In the 1970s, before home computers made this much easier, my father Dr James F. Coppedge ran an experiment to look for meaningful sequences in random Scrabble letters selected with replacement from a full set of 26 letters of the English alphabet. Out of 30,000 trials, ABC appeared only once. There were over a thousand 3-letter meaningful combinations, 139 with 4 letters, 17 with 5 letters, 3 with 6 letters, and 1 with 7 letters, illustrating the exponential drop with sequence length. He calculated that to find the phrase “the theory of evolution” this way, using a machine that could draw a billion letters per second, the time required for any attempt on average would vastly exceed the age of the universe: 26 quadrillion years.
3. Notice the difference between random and designed in this quote from the Conclusion: “Our results show that the inclusion of repeats enriches the fraction of otherwise random sequences that can adopt folded conformations. … On the other hand, beyond a repeat length of 30 residues, the frequency of observing predicted folded structures decreased sharply, reaching 0.001% for sequences lacking repeats. Also, these structures were all globular. Evidently, without the imposition of a sequence repeat, the probability of adopting a repeating conformational form becomes very low, and only more asymmetric arrangements typical of the domains seen in globular proteins are observed.… However, when we attempted to express these randomly generated sequences we found that their expression level was low, and they failed to fold into the desired structures as assessed by CD [circular dichroism spectroscopy]. Thus, we conclude that these were adversarial sequences, and the probability of finding sequences that fold into native structures at random is less than one in a million. However, we could use the predicted structures as starting points for design of native-like proteins using the sequence design module of Chroma to improve their sequences. The redesigned sequences expressed well, and gave highly stable and tertiary structures as assessed by temperature-dependent CD spectroscopy.”
4. Meyer, Signature in the Cell (2009), pp. 210-212.
5. Similar question-begging claims are often found in discussions of duplicated genes, where the duplicate gene is said to be “freed up” to evolve a new function.
6. E.g., “New tertiary structures might, however, confer advantages for applications or environments not currently found in nature.”
7. Popper and Eccles, The Self and Its Brain, 1977, sec. 26. “Promissory materialism is a peculiar theory. It consists, essentially, of a historical (or historicist) prophecy about the future results of brain research and of their impact. This prophecy is baseless. No attempt is made to base it upon a survey of recent brain research. The opinion of researchers who, like Wilder Penfield, started as identity theorists, but ended as dualists (see Penfield [1975], pp. 104f.) is ignored. No attempt is made to resolve the difficulties of materialism by argument. No alternatives to materialism are even considered.”